Cloud Fusion has three defining attributes: • It delivers advanced integration of multiple clouds – interconnecting the clouds of today to deliver the “cloud of clouds” for business • It enables actionable insights from big data, analyzed and aggregated to support real-time business navigation • It delivers extended services from the cloud for distributed applications which handle mobile devices/sensors and provides all necessary resources/services in an on-demand self-service model.

1. Cloud Fusion Concept
 Yoshitaka Sakashita  Kuniharu Takayama  Akihiko Matsuo
 Hidetoshi Kurihara
Cloud is expected to develop from a single-purpose cloud to a hybrid cloud that
links clouds or existing systems, or to a fusion of two or more clouds in the future.
Fujitsu Laboratories named this advanced form of coordination “Cloud Fusion” at
the start of 2010. This paper explains the aim of this coordination and the direction
in which research should head. It goes on to describe the relationship between
Cloud Fusion and Fujitsu’s and Fujitsu Laboratories’ vision—enabling Human-
Centric Intelligent Society. It describes the five pillars of research on Cloud Fusion
and its outline. In particular, it introduces details about the development and
execution environment that is one of the pillars.
1. Introduction 2010.note) i We have conducted research while
Fujitsu Laboratories started working on closely working with the product development
cloud computing (hereafter “cloud”) together division and FUJITSU Family Association/LS
with Fujitsu in the spring of 2007 and started up Research Committee, a customers’ research
the present Cloud Computing Research Center committee, from the beginning. In the process
in April 2009. This organization is an aggregate of studying what to realize with clouds and
of groups engaged in cloud-related research such communicating with customers, we became
as infrastructure control, system management aware that technology capable of linking existing
and development and execution environment. It systems and clouds was necessary. Hence, we
has been launched as a center with the nature started to work on achieving a hybrid cloud
of a cross-laboratory organization. Fujitsu and around the summer of 2009. With this idea still
Fujitsu Laboratories had been researching promoted as Fujitsu’s cloud strategy, we have
themes such as grid computing and virtual proposed the optimum system construction by
system control since before the launch of the linking public and private clouds and existing
center. As an initial achievement, we developed systems and offered products. Subsequently, we
a virtual system control engine, which provides have proposed various types of clouds to give rise
the core of a cloud system.1) The engine was to healthcare, local government and other clouds.
adopted as an in-house practice system of Under the circumstances, Fujitsu Laboratories
the On-demand Virtual System Service. It is felt it was necessary to develop a system linked
capable of providing an infrastructure for the with various clouds and existing systems, and
development of necessary systems such as user named this form of coordination “Cloud Fusion”
management and billing as well as operation and note) i In June 2011, the On-demand Virtual
service design. Moreover, it contributed to the System Service was named as Fujitsu
subsequent start of commercial service in October Global Cloud Platform “FGCP/S5” service
in the Japanese market and Global Cloud
Platform (GCP) service in overseas markets.
FUJITSU Sci. Tech. J., Vol. 48, No. 2, pp. 143–150 (April 2012) 143

2. Y. Sakashita et al.: Cloud Fusion Concept
at the start of 2010. described above, we estimated that a system for
This paper describes the aims and overall linking and mutually using private and public
concept of this Cloud Fusion. clouds would be important in terms of business
as well. We aimed to link various clouds together
2. Aims of Cloud Fusion by Cloud Fusion to link and share information,
In the cloud market, the public cloud and thereby expand the fields of information and
business and software as a service (SaaS) communications technology (ICT) application
business based on it were the mainstream in and create a new market. A similar concept
the initial phase, and we now expect customer- is referred to as “inter-cloud” by the Smart
specific private cloud services will develop. IDC Cloud Study Group of the Ministry of Internal
Japan released a forecast on April 7, 2011 saying Affairs and Communications (MIC).2) While
that the scale of the software market for public similar concepts have been recently announced
clouds and that for private clouds will be almost by other companies as well, we believe that
equal at 145 billion yen in 2012, and the latter Fujitsu Laboratories has taken the lead in this
will overtake the former after that. Meanwhile, research. Furthermore, Fujitsu Laboratories
a form of sharing a cloud by an industry or proposes a way to eliminate companies’ anxiety
community has appeared. At present, we are of being locked in to a specific cloud by using
conducting research on media and engineering Cloud Fusion to link different types of clouds.
clouds as a service in addition to healthcare and As technologies required for realizing these aims
local government clouds. In addition, we are and ideas, we are intensively engaged in R&D
currently linking clouds by using sensors and under the following five themes.
mobile phone handsets as input devices, which • Secure data connection
may be referred to as the Internet of Things. • Big data parallel distributed processing
Fujitsu and Fujitsu Laboratories set forth • Cloud development and execution
a vision of Human-Centric Intelligent Society in environment
the spring of 2010 and have since been working • Cloud mangement
for the construction of a cloud environment • Managed network
that supports this vision. Based on the trends Figure 1 shows the background, concepts
From hybrid cloud to Cloud Fusion, multiple
clouds freely combined for use
Highly-integrated
cloud services
Community
Hybrid cloud Industry Medical
care
Enterprise
Cloud
Media
Engineering
Single cloud Corporate
information
center
Figure 1
Fujitsu Laboratories’ approach to cloud.
Figure 1
144 FUJITSU Sci. Tech. J., Vol. 48, No. 2 (April 2012)
Fujitsu Laboratories’ approach to cloud.

3. Y. Sakashita et al.: Cloud Fusion Concept
and major research themes. The following entire system. For that purpose, aside from the
sections describe the R&D for realizing Cloud research group for Cloud Fusion, we have other
Fusion with the focus on the development and groups such as one engaged in research on Green
execution environment. For other themes, see Internet Data Center (IDC), for example, which
the respective papers contained in this special puts together servers, storage, networks and
issue. facilities. Above Cloud Fusion, there is also a
group working on the technologies for enabling
3. Technologies for realizing Human-Centric Intelligent Society, which is
Cloud Fusion Fujitsu’s and Fujitsu Laboratories’ vision. In
An overview of the technologies for realizing this field of research, Cloud Fusion takes charge
Cloud Fusion is shown in Figure 2. of the development and execution environment
Cloud is a synthesis of technologies for high-speed accumulation and analysis of
widely ranging from infrastructures such as various pieces of event data and other big data.
hardware and networks to technologies that Human-Centric Intelligent Society supports
use them to make contributions as business data collection in the actual field and analysis,
and social infrastructure. Accordingly, Fujitsu prediction and proposal for use in the field.
Laboratories does not only conduct world-class The Cloud Fusion group has defined five
R&D in the respective fields of these technologies pillars of R&D as mentioned in the previous
but also links other companies’ technologies and section. The following outlines the respective
open source materials. It does so by vertical pillars of research.
integration in an attempt to sophisticate the 1) Secure data connection
Human-Centric Intelligent Society
Fusion
Cloud Fusion technologies Management
Cross-industry cloud
Secure data
connection
Other company’s cloud Big data PaaS management
parallel distributed Information hiding
processing
Information traceability Infrastructure
Linking Large-scale parallel
integrated visualization/
Development and distributed
Industry cloud Logic movement diagnosis
processing
execution environment between clouds
Group company cloud Development simplification Volume event Predictive fault
processing detection
Data analysis
application
development and Managed
Single cloud integration environment Distributed KVS
network
Other company’s SaaS Application execution Volume batch
environment management processing Increase of throughput
platform
of communication
Public cloud Enterprise mashup between DCs, etc.
technology
Private cloud Test automation/
efficiency improvement Virtual platform control/construction
Existing system
Green IDC Cloud management standardization
Figure 2
Fujitsu Laboratories’ overall technology for Cloud Fusion.
Figure 2
Fujitsu Laboratories’ overall technology for Cloud Fusion.
FUJITSU Sci. Tech. J., Vol. 48, No. 2 (April 2012) 145

4. Y. Sakashita et al.: Cloud Fusion Concept
Security, which is said to be a major of IT management, aiming for its standardization.
hindering factor to the application of cloud, may In addition, we have commercialized part of the
have a significant impact if information is leaked technology to comprehensively visualize and
and a system to allow use with more security diagnose the performance of network and server
than before is required. We want customers to infrastructure at the end of 2010. Since the
entrust Fujitsu’s cloud with their information beginning of 2011, we have been aiming to apply
without anxiety. Hence, we aim to be able to predictive fault detection technology to the real
hide information between the existing systems environment. To this end, we have been moving
and clouds and to identify how information has ahead with new research on PaaS management
been used (this function will be commercialized in consideration of application life cycles.
in the middle of FY 2012). In addition, we need 5) Managed network
to be able to process data without handing them In this age of cloud, we believe networks will
to a cloud. Therefore, we have researched and experience a dramatic increase in data volume,
developed a technology to have SaaS applications and this has given rise to demand for research
downloaded to the own center and impose a to address various challenges. For example,
restriction to prevent the applications from research on the performance and management
sending information outside. technology of a network between clouds and
2) Big data parallel distributed processing between clouds and terminals is necessary and a
The essential aim of Cloud Fusion is to new perspective is required.
link and share information to expand the fields
of ICT application and create a new market. To 4. Overview of development and
do this, big data processing is a key technology. execution environment
We are basing our work on technologies such as The essence of cloud is aggregation and
the key value store (KVS) technology, suitable distribution. That is, cloud is intended to
for storing big data, and open source Hadoop, aggregate on a large scale customers’ systems
capable of quickly processing big data. And we and data to realize economies of scale and
are investigating even more advanced parallel their distributed processing by commodity
distributed processing for close linkage with infrastructure (such as servers, storage and
Human-Centric Intelligent Society. networks) to realize cost efficiency.
3) Cloud development and execution In terms of execution environment,
environment aggregating systems on a cloud provides cost
This involves research on technology to and quality benefits. For example, it eliminates
allow customers to easily connect clouds with the need for infrastructure on the customer’s
existing systems and big data analysis platforms. site, makes systems available for use only when
The following section gives the details. and as much as required, and acts as measures
4) Cloud management for disaster recovery by using a backup cloud.
Up to now, we have conducted research Meanwhile, aggregating data on a cloud also
on the efficient control and management of provides cost, quality and efficiency benefits in
virtualization systems. We have also proposed terms of development environment. For instance,
the management interface for On-demand it centralizes the development environment
Virtual System Service. It became commercially and makes it easy to ensure governance on
available in October 2010, and we proposed it to the development processes and results. With
the Distributed Management Task Force (DMTF), this situation in the background, software
an industry group promoting the standardization development and execution environment on
146 FUJITSU Sci. Tech. J., Vol. 48, No. 2 (April 2012)

5. Y. Sakashita et al.: Cloud Fusion Concept
a cloud, or the so-called platform as a service fusion of various application systems will be
(PaaS), has recently been evolving and becoming important including the existing and new
widespread. areas. With general Web applications, linking is
It must be noted that implementing existing roughly classified into linking in the front end in
business applications on a cloud has a scale-out charge of interaction with the user and that in
problem. With the conventional infrastructure, the back end, which takes charge of processing
business applications with large amounts of and interchange with databases according to
accesses and processing are often built by using the access from the front end. One well-known
high-performance servers and databases, which technology of the former is mashup. For the
is called scale-up. On the other hand, clouds are latter, service-oriented architecture (SOA) is a
composed of large numbers of similar commodity popular technology. One important challenge in
servers and databases and cannot be scaled up. such linking is how to effectively use the existing
Accordingly, scale-out is necessary, in which systems. Accordingly, we have developed
processing is distributed among many servers WebAPI creation assistance technology,
and databases. Relational databases (RDBs), which semi-automatically extracts application
which are typical conventional databases, programming interfaces (APIs) from the existing
are good at complicated data retrieval and systems to allow them to be mashed up with
aggregation. However, they take their processing other applications.
steps one at a time so as to maintain consistency, Of these, big data analysis PaaS and
which makes them unsuitable for distributing WebAPI creation assistance technology will be
multiple accesses for parallel processing. In explained in more detail in the following sections.
clouds, on the other hand, new types of data
storage are used such as KVS, which is capable 5. Big data analysis PaaS
of high-speed parallel distributed processing of From now on, it will be important to make
a simple data structure with sets of keys and use of data gathered in large volumes from
values. For this reason, we have developed sensors and terminals to propose new services
optimum deployment technology for data storage. and solutions that could not be readily realized
It automatically selects between multiple types with the conventional ICT. In the analysis
of data storage such as RDB and distributed KVS and development environment based on Cloud
in a cloud system according to the characteristics Fusion, we will need to create new services
of the data or accesses handled. In this way, it together with customers by combining data
appropriately switches data storage accesses across the boundaries of business categories and
from Web applications.3),4) industries starting from the big data owned by
This paper has so far covered the existing customers.
areas of business. New business areas that take However, analyzing big data and developing
advantage of big data such as sensor data and event-driven services requires skills to widely
business logs are being launched. To analyze master and implement various processing
large amounts of data and develop event-driven technologies including data parallel distributed
services, we need a system capable of easily and processing (such as Hadoop), event processing
comprehensively analyzing and developing batch (such as Esper) and analysis processing (such
analysis processing of big data and real-time as Mahout). An analysis and development
processing of event data. We have developed a environment that comprehensively supports
big data analysis PaaS for that purpose. them has not been realized.
In the future, we expect that linking and To address this need, we have test-built the
FUJITSU Sci. Tech. J., Vol. 48, No. 2 (April 2012) 147

6. Y. Sakashita et al.: Cloud Fusion Concept
environment shown in Figure 3 that supports We have taken a real case of business as a
the cycle of service planning by data analysis reference to develop cluster analysis processing
[2) in Figure 3], simplified development of and coupon issuing application and evaluated
service applications [3) in Figure 3] and service this environment. We have confirmed that with
evaluation [4) in Figure 3]. In this environment, this environment it is possible to reduce the
batch analysis processing of big data and real- period of time taken to provide the service, which
time processing of event data can be handled in an was conventionally two months, to two weeks.
integrated manner. As an interface, a data flow
diagram (DFD) editor environment (HTML5) is 6. WebAPI and mashup
provided [1) in Figure 3]. The architecture allows technology
the defined DFD to be configured as a program More than one business system is operated
according to Service Component Architecture in an enterprise and a user often needs to
(SCA), a standard component specification, for use multiple systems to perform a series of
deployment and execution. operations. Along with the expansion of cloud,
The data analysis personnel can easily companies now need to link with applications
execute analysis and event processing without provided by external clouds in addition to those
needing knowledge about the individual systems. For this reason, we have developed
implementation technologies, simply by WebAPI creation assistance technology to call
deploying input/output (I/O) data and processing existing systems as Web services without making
components as a DFD and providing execution changes to allow linking by mashup.
parameters. To realize this system, we have Mashup is a technique of combining
developed as a differentiating technology a new multiple Web services published on the Internet
function to convert components into the optimum to build Web applications with new value,
implementation component according to the as exemplified by Google Maps. Application
types of processing and data. developers can easily use functions such as data
Feedback of points of improvement
Evaluation Planning
4) Service effectiveness analysis 1) Analysis process description by DFD 2) Static analysis of large-scale data
Multidimensional Cluster
Effect Service analysis analysis
measurement use Accumulated
data Hive Mahout event data
RDB Hadoop HDFS/MapReduce
HTML5
Service
customization
Simplified Service development
development reflecting business rules
3) Real-time processing by CEP
Service operation
Service
Event Filter Matching notification
Queue CEP engine Queue
CEP: Complex event processing
Figure 3
Outline of PaaS for large-scale data analysis.
Figure 3
148 FUJITSU Sci. Tech. J., Vol. 48, No. 2 (April 2012)
Outline of PaaS for large-scale data analysis.

7. Y. Sakashita et al.: Cloud Fusion Concept
acquisition and screen display/operation simply them to be used from mashup applications. With
by executing calls according to the APIs offered this technology, the user’s screen operations on
by Web services. It may be possible to apply this the system are reproduced to obtain the results
mashup technique to enterprise systems to use of processing from the screen, thereby adding
the functions of existing systems and applications WebAPIs without making changes to existing
that meet the needs of user departments. If such systems. Figure 4 shows an overall view of this
an environment can be built easily then higher technique.
operations efficiency can be expected.5) To create an HTTP request queue as
To use system functions by mashup, a reproduction of screen operations, this
however, the functions must be provided as APIs technology has the user execute the operations
and callable from a different system. Existing of the applications concerned and records data
systems in enterprises are often not provided sent and received between the browser and
with such APIs and are unavailable from outside Web applications. This eliminates the need
as components. Making changes to systems to develop a program for generating a request
to add APIs or data-linking function requires queue and operations that need many screens
many person-hours for development and testing. before producing results can be easily turned
This has made it difficult up to now to use the into WebAPIs. In addition, it has a function to
functions of existing systems by mashup. edit the recorded request queue for extraction of
To deal with this challenge, we have parameters to be provided from outside such as
developed WebAPI creation assistance the user ID, password and search key. Hence,
technology. It adds APIs without making authentication information essential to business
changes to existing Web applications to allow systems can be provided by mashup applications.
Developer Enterprise
mashup proxy
1) Execute operations
HTML
Business application
JSON
Communication/response recorder
3) Edit and provide components 2) Record communication log
Communication/response records
Process
SaaS application
executed
GET / HTTP/1.0 according to
GET /login.cgi the records
HTTP/1.0
POST /search
User HTTP/1.0
Key=XXX&id=YYY
Generalized business component
Present development
4)
Mashup by using
components
Figure 4
Overall support technology for making WebAPI.
Figure 4
FUJITSU Sci. Tech. J., Vol. 48, No. 2 (April 2012)
Overall support technology for making WebAPI. 149

8. Y. Sakashita et al.: Cloud Fusion Concept
In a trial using this technology, we have Fujitsu Laboratories intends to continue to lead
confirmed that WebAPI can be created with one- the development of cloud technology so as to help
tenth the person-hours as compared with adding create and pioneer new business.
WebAPI to an existing system by manually
creating a request queue. References
1) M. Kishimoto: Fujitsu’s Trusted-Service
Platform. (in Japanese), Cloud Computing
7. Conclusion Technology to Grasp the Other Side of the World
Fujitsu Laboratories’ approach to the area of Clouds. ASCII, 2009, pp. 58–65.
2) MIC: Smart Cloud Study Group Report—Smart
of cloud presented here is based on the vision Cloud Strategy—(May 2010).
called Cloud Fusion. At the beginning of the http://www.soumu.go.jp/main_sosiki/
joho_tsusin/eng/councilreport/pdf/
development of On-demand Virtual System 100517_1.pdf
Service in October 2010, it was referred to as the 3) M. Adachi et al.: Fujitsu’s Approach to the Age
of Cloud—Development of Construction and
Service-Oriented Platform. We thought of a “site” Operation Technology for a Cloud System. (in
where information is gathered to conduct new Japanese), ASCII technologies, June Issue, pp.
62–69 (2010).
communication and business as a “platform.” We 4) Y. Mizobuchi et al.: Automatic Optimized Use of
have conducted research based on this idea since Multiple Datastores in Cloud Computing Era.
(in Japanese), IPSJ Special Interest Group on
before the name “cloud” became popular and we Software Engineering Report, Vol. 2010-SE-170,
think that the essence of Cloud Fusion is the No. 14, pp. 1–8 (2010).
5) A. Matsuo et al.: LivePoplet: Technology That
same. In the future, cloud is expected to change Enables Mashup of Existing Applications.
the business, social infrastructure and lifestyles Fujitsu Sci. Tech. J., Vol. 45, No. 3, pp. 304–312
(2009).
of people in the world more than we can imagine.
Yoshitaka Sakashita Akihiko Matsuo
Fujitsu Laboratories Ltd. Fujitsu Laboratories Ltd.
Mr. Sakashita is currently engaged in Mr. Matsuo is currently engaged
research strategy and promotion for in research and development of
entire cloud computing. inter-cloud linking technology and
technology to improve efficiency of
software maintenance.
Kuniharu Takayama Hidetoshi Kurihara
Fujitsu Laboratories Ltd. Fujitsu Laboratories Ltd.
Mr. Takayama is currently engaged Mr. Kurihara is currently engaged in
in research and development of research and development of software
application development and execution development environment and
envi-ronment in cloud. up-stream process technology.
150 FUJITSU Sci. Tech. J., Vol. 48, No. 2 (April 2012)